Jeremy Bentham is widely known for his political theory of utilitarianism, but far less recognized for his sophisticated reflections on language. As part of the broader Enlightenment project, Bentham—like many philosophers of his time—sought to refine language so that it could serve as a clear and adequate medium for describing the world.

The ambition was both simple and radical: to strip language of its imaginative distortions in order to eliminate sources of confusion and error. These were not merely epistemological problems, but also sources of social harm and injustice. Bentham’s project was therefore both theoretical and practical.

If we seek a consistent and reliable ontology, Bentham believed, we must first correct the linguistic structures through which that ontology is expressed. As long as language remains imprecise, our understanding of the world will remain unstable—shaped not only by experience, but also by layers of metaphor, abstraction, and imagination.

One might paraphrase Ludwig Wittgenstein here: the limits—or even the distortions—of our language become the limits—or distortions—of our world.

When social relations are structured upon such unstable or misleading ontological premises, the consequences extend beyond mere confusion. They become embedded in institutions, norms, and practices. In this way, imprecise language does not simply misdescribe reality—it participates in shaping it. What begins as a linguistic shortcut can solidify into a framework that guides judgment and action.

It is therefore less surprising than it might first appear that contemporary social and political realities continue to rely on constructs that are, in Bentham’s sense, fictitious. The issue is not that such fictions exist—they are unavoidable—but that they often operate without being recognized as such. When this happens, their influence becomes indistinguishable from necessity, and their consequences—sometimes unjust—are absorbed into the structure of everyday life.

Bentham’s analysis led to a surprising and somewhat paradoxical conclusion. Language cannot, in fact, be fully purified. It inevitably relies on what he called fictions: entities that do not exist in the physical world, but are nevertheless indispensable for thought and communication—that is, for the world as it appears to us. Concepts such as obligation, right, power, or even corporation do not correspond to tangible objects; they are linguistic constructions that allow us to, create, organize and navigate complex social reality.

The problem, however, runs deeper than mere non-existence. These fictions are not inert. They shape reasoning, justify actions, and produce real-world consequences. We act as if they were real—and in doing so, they acquire a kind of practical reality. In this sense, fiction does not merely obscure truth; it participates in its construction.

Even if we set aside the many technical derivatives — such as “machine learning” or “predictive modeling” — the term artificial intelligence itself is deeply misleading. None of these systems are intelligent, regardless of the sophistication or usefulness of the outputs they produce. The designation does not describe an underlying property of the system; rather, it introduces a fiction that suggests the presence of cognition, understanding, or agency. What we are dealing with are highly advanced systems of statistical inference — tools that identify patterns, optimize outputs, and generate responses based on probabilistic structures. Yet the language used to describe them systematically exceeds their actual capacities. In Bentham’s terms, “intelligence” in this context functions not as a real entity, but as a linguistic construct — one that organizes perception while simultaneously distorting it.

Imaginary, Symbolic, Real: A Lacanian Anatomy of AI

These insights find their most telling contemporary echo in the later work of Jacques Lacan, whose structural analysis of subjectivity operates through three interconnected registers: the Imaginary, the Symbolic, and the Real. Lacan suggested that truth itself has the structure of a fiction—but this claim becomes clearer when we examine how all three registers are at work in our relation to AI systems.

The Imaginary register is the most immediately visible. This is the domain of images, identifications, and mirror relations—where we misrecognize ourselves and others by projecting familiar forms onto unfamiliar entities. AI systems are heavily imaginarized: we assign them names, read intention into their responses, and experience a form of recognition in their outputs. This goes far beyond what is epistemologically or socially comfortable. People treat language models as intimate friends, partners, confidants, advisors, medical experts, and coworkers. They ask a statistical model — one trained to predict the most probable next token in a given context — to supply the intimate and social dimensions of their lives. In doing so, they avoid the actual social context that is a crucial constituent of what it means to belong to human community. The fiction of intelligence begins here, as an imaginary capture — a reflection mistaken for a presence. The long-term social consequences of this dynamic are potentially the most disturbing, not least because we are largely failing to recognize them as such. We are beginning to act as though a surrogate for society is sufficient for society to exist. Deprived of social responsibility, genuine empathy, and the friction of real interdependence, we are learning to be satisfied with the practical residue — what Bentham might call the practical reality — of a statistical language prediction model. It seems trivial. It is not.

The Symbolic register is where Bentham’s fictions properly reside. This is the order of language, law, and institutional structure — the domain of fictitious entities that produce real effects precisely because they are embedded in shared frameworks. When “artificial intelligence” enters legal documents, regulatory systems, or corporate structures, it ceases to be merely an image and becomes a symbolic fiction: one that generates obligations, distributes responsibility, and shapes judgment, regardless of whether it corresponds to any underlying reality. Bentham and Lacan converge here most directly — both recognized that language does not simply describe a pre-existing world, but participates in constructing it. Where the Imaginary fiction is intimate and personal — a misrecognition that happens in the individual encounter — the Symbolic fiction is collective and structural. It does not announce itself. It accumulates quietly inside institutions, redistributing agency and responsibility until the social reality it has shaped becomes difficult to distinguish from the one we chose. That this process is accelerating should surprise no one: in an order where economic logic governs institutional adoption, efficiency does not wait for understanding. It is economic logic, more than epistemic ambition, that is driving the rapid penetration of these systems into the structures that define our sociality and articulate our mutual obligations. When the military, lawyers, teachers, and municipal workers adopt them — not because the fiction has been examined, but because the efficiency gain is real — the symbolic embedding precedes any honest reckoning with what is actually being embedded.

The Real, however, is the most unsettling register. In Lacan, the Real is not simply “reality,” but that which resists symbolization—what cannot be fully captured by either image or language. What these systems actually are—vast engines of statistical inference operating through high-dimensional computation—remains Real in this sense. Every attempt to describe them as “intelligent,” “learning,” or “understanding” overlays this reality with symbolic and imaginary constructs that fail to capture their operation. The gap between the fiction and the system itself is not a temporary limitation—it is structural. In Lacanian terms, our current discourse around AI has the structure of a symptom: a formation that simultaneously organizes experience and conceals what it cannot absorb. We do not simply misname these systems out of ignorance or convenience — we misname them because the accurate name would force a confrontation with what the fiction is designed to avoid. The longer the symptom persists unexamined, the more the Real it papers over will insist — not through revelation, but through the accumulation of consequences we will find increasingly difficult to account for.

This is why Lacan’s claim that truth has the structure of a fiction is more than a paradox when applied to artificial intelligence. Here, the three registers become visibly misaligned: we imagine a mind, we legislate an agent, but the Real of the system escapes both. What Bentham uncovered at the level of legal and philosophical language thus anticipates something broader: that our engagement with such systems is always mediated by constructs that are, strictly speaking, not real.

The Fiction Beyond Recognition

Bentham’s warning was never against fictions as such. He understood that language cannot function without them — that obligation, right, and corporation are indispensable constructs, not errors to be eliminated. His concern was more precise and more troubling: It was with fictions that exist beyond recognition. Against constructs that conceal their own constructedness so completely that they begin to pass as necessity.

Artificial intelligence, in this sense, is not simply a misleading term. It is a fiction so thoroughly naturalized that we have largely lost the capacity to see it as one. What began as a naming decision — ambitious, perhaps convenient, certainly consequential — has accumulated across three registers simultaneously. It has captured the imagination, reorganized the symbolic order of institutions and law, and in doing so has built a wall between us and the Real of what these systems actually are. We imagine a mind. We legislate an agent. And the statistical engine underneath continues its operation, indifferent to both.

What matters, then, is not what these systems are — that question, though important, remains largely confined to technical discourse. What matters is what we have agreed to call them, and what that agreement permits us to do: to automate judgment without attributing responsibility, to replace social bonds with their statistical simulacra, to embed a fiction so deeply into institutional structure that challenging it begins to feel like challenging reality itself.

The fiction of intelligence would be relatively harmless if recognized as one. Bentham knew this. Fictions, acknowledged, can be used carefully and revised when necessary. It is the naturalization of the fiction — its disappearance into the background of the obvious — that transforms a linguistic shortcut into a framework for action, and eventually into a structure of consequences that no one chose and everyone inherits.

We are not at the end of this process. We are somewhere near its acceleration. The question Bentham’s work leaves us with is not whether we can do without fictions — we cannot — but whether we retain the critical capacity to recognize them for what they are before the Real they conceal insists in ways we are no longer equipped to understand.